# context-bridge [Health: Active]

**Category:** 🗣️ Conversational AI  
**Repository:** https://github.com/prateekg7/Context-Bridge  
**GitHub Stars:** 3  
**npm Downloads (last month):** 159  
**Views:** 1  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/context-bridge

## Description
Context portability MCP server for compressing, formatting, and resuming AI conversation handoffs.

## Tools
Capabilities this server exposes over MCP:

- **compress_context** — Compresses a raw transcript into a structured context block.
- **format_for_target** — Formats a compressed context block for a destination platform.
- **resume_context** — Builds a destination-ready landing package from a compressed context block.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "context-bridge": {
    "command": "npx",
    "args": ["-y","@contextbridge_ai/mcp"],
    "env": {
      "OPENAI_API_KEY": "",
      "OPENAI_MODEL": "",
      "OPENAI_BASE_URL": "",
      "GOOGLE_API_KEY": "",
      "GROQ_API_KEY": "",
      "CONTEXTBRIDGE_API_KEY": "",
      "CONTEXTBRIDGE_HOSTED_URL": "",
      "CONTEXTBRIDGE_HOSTED_MODEL": ""
    }
  }
}
```

**Requires environment variables:** `OPENAI_API_KEY`, `OPENAI_MODEL`, `OPENAI_BASE_URL`, `GOOGLE_API_KEY`, `GROQ_API_KEY`, `CONTEXTBRIDGE_API_KEY`, `CONTEXTBRIDGE_HOSTED_URL`, `CONTEXTBRIDGE_HOSTED_MODEL` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What context-bridge does

The context-bridge MCP server turns a raw conversation transcript into a structured context block for continuation elsewhere. The block is intended to preserve the parts of a working session that a second AI needs: the user’s objective, decisions already confirmed, current progress, unresolved topics, referenced artifacts, constraints, recent changes in direction, and a suggested next action.

After compression, the server can adapt that block for Claude, ChatGPT, or Gemini. It can also create a landing package designed to help the destination model resume the work without requiring the user to repeat the original discussion.

The published package is the stdio MCP server distributed through npm as `@contextbridge_ai/mcp`. The browser extension is maintained in the same repository but is not included in this MCP Registry publishing path.

## How it works

The server exposes three MCP tools:

- `compress_context` accepts a required transcript and produces a validated structured context block. Optional arguments can select a provider, model, API key, base URL, or hosted-lane key.
- `format_for_target` accepts a context block and a required destination of `claude`, `chatgpt`, or `gemini`, then formats the handoff for that platform.
- `resume_context` accepts the same kind of context block and destination, then creates a destination-ready landing package.

Compression can use an OpenAI Responses lane, Google, Groq, a custom provider, or a hosted endpoint. Provider selection can be supplied to the tool, while environment variables provide defaults. The documented priority uses an explicit compression key first, then `OPENAI_API_KEY`, then the hosted URL configuration.

The context-bridge MCP server communicates through stdio, so an MCP host starts it as a local process.

## Setup and configuration

Install the package through an MCP client configuration and provide one supported inference route. An OpenAI setup can define `OPENAI_API_KEY`, `OPENAI_MODEL`, and `OPENAI_BASE_URL`. Google, Groq, and custom OpenAI-compatible endpoints have corresponding provider settings. The hosted route uses `CONTEXTBRIDGE_HOSTED_URL` and `CONTEXTBRIDGE_HOSTED_MODEL`, with an optional `CONTEXTBRIDGE_HOSTED_API_KEY`.

The tool arguments can override provider, model, API key, and base URL for an individual compression request. A hosted configuration is intended as a no-key path for users who do not supply their own provider credentials, although the hosted endpoint itself may optionally require an authentication key.

The project is licensed under MIT. The README provides no claim that transcripts are stored, encrypted, or processed locally; operators should evaluate the selected provider or hosted endpoint before sending conversation content.

_Full upstream README: https://allmcps.com/mcp/context-bridge/readme_